Zian Zeng← Back to websiteDownload PDF ↓

Curriculum vitae

Zian Zeng

M.S. Computer Science · University of Maryland, College Park

Reliable language-model agents; Tool use and action admission; Agent recovery and evaluation; Multi-agent collaboration; Structured reasoning.

Education

M.S. in Computer Science

Aug 2025–May 2027 (expected)

University of Maryland, College Park

B.S. in Computer Science

Aug 2020–May 2025

University of Hawaiʻi at Mānoa

Research experience

University of Maryland, College Park

First author and sole student author on five current manuscripts; led implementation, experimental design, analysis, writing, and revisions. Research with Professor Jordan Boyd-Graber (CLIP Lab), Professor Furong Huang (Furong Lab), and Professor Ashok Agrawala (MIND Lab) spans action admission, agent recovery, and evaluation validity, alongside collaborative work on RoleFlow.

University of Southern California · GLAMOR Lab

Contributed data-augmentation, decoding, and embedding experiments to neuro-symbolic American Sign Language modeling. Coauthored the NAACL 2025 Findings study combining 11 linguistic sources and three video tasks; the collaborative models achieved 91% isolated-sign recognition accuracy.

University of Missouri · VIMAN Lab

With graduate mentor Roland Oruche, contributed implementation and experiments to three papers on multi-hop question answering and conversational advising. Implemented four prompting/retrieval comparisons and worked on the Vidura / KnowCOVID-19 scientific-literature platform.

University of Hawaiʻi at Mānoa · Hawaii Digital Health Lab

Prepared first-author research comparing CNN and Vision Transformer models for autism spectrum disorder classification on 547 eye-tracking images from 59 participants. Compared saliency maps, Grad-CAM++, LIME, and attention maps; also contributed to human-in-the-loop crowdsourcing and multimodal autism/ADHD research.

Publications

EACL 2027

CallAdmit: Diagnosing and Efficiently Repairing Over-Abstention in Tool-Call Admission with Fixed-Proposal Certificates

Zian Zeng, Jordan Boyd-Graber (2026).

NAACL 2027

Restoration Is Not Neutrality: Auditing Visible Budget Controls in Agent Recovery

Zian Zeng, Furong Huang (2026).

COLING 2027

Correct Statuses, Wrong Targets: Truthful Goal Feedback Can Misdirect Stateful Repair Agents

Zian Zeng, Furong Huang (2026).

NAACL 2027

EffectTwin: Separating Recovery Limits from Output-Protocol Effects in Tool-Using Language Models

Zian Zeng, Ashok Agrawala (2026).

NAACL 2027

The Agent or the World? Verifier Inflation in Persistent-World Agent Evaluation

Zian Zeng, Ashok Agrawala (2026).

IEEE Big Data 2024

Chain-of-Query Prompting for Efficient Small Language Models in Multi-Hop Open-Domain Question Answering

Rithika Akula, Roland Oruche, Zian Zeng, Yuanxun Zhang, Marcos Zampieri, Prasad Calyam (2024).

Professional experience

Honolulu Fire Department

Built an incident/performance dashboard integrating three systems—CAD, RMS, and GIS—through C# APIs, SQL Server, and Angular/ASP.NET Core. Developed scheduled incident-alert reports and Python/Outlook analytics for media inquiries and response times.

Blue Planet Energy

Contributed monitoring-dashboard and control-interface prototypes for high-voltage solar-battery systems using Python, Raspberry Pi, and PLCs. Maintained cloud-based inventory and sales records and continued in a part-time role following the internship.

Teaching and mentoring

Windward Community College

Led weekly Java/introductory computer science study sessions, group learning activities, and office hours for 40+ students across six semesters.

Windward Community College

Provided individualized mathematics tutoring through Calculus II for 20+ students across three semesters; independently operated the summer math lab for 2.5 months.

Selected projects and leadership

ORBIT

Led a Flutter/Dart student-support prototype with five specialist agents, shared context, and a Node.js data proxy. Evaluated the prototype with 40 synthetic student profiles. Led the UMD xFoundry entry and 75-page business plan, guiding teammates’ writing and chart creation around student-controlled support, institutional adoption, and a proposed campus pilot.

Hybrid Edge–Cloud Agentic Reasoning

Global competition across five host cities. Designed routing across deterministic logic, on-device FunctionGemma, and Gemini cloud fallback. Ranked #2 on the Washington, DC leaderboard at competition close.

MUIR

Exploring marginal-utility routing for retrieval, checking, repair, verification, and stopping in tool-using agents.

Quiz App

Led technical development and requirements for a three-person University of Hawaiʻi at Mānoa capstone team sponsored by Professor Edoardo Biagioni. Built the Flutter/Dart student client for a classroom quiz system with a companion instructor desktop app, including photo-based attendance capture, persistent profiles, and real-time submissions. Adapted the student client from Bluetooth experiments to LAN-based WebSockets with a shared connection manager across screens, working through Android permission and connection-handling issues.

Saber Quest: Echoes of the Fallen

Originated the game concept and led project framing, resource selection, and task allocation for a two-person team with Feiyi Chen. Co-developed the VR action game using Unreal Engine 4.26 and Blueprints; implemented lightsaber effects and activation behavior, controller interactions, menus, player health, and damage/death feedback. Resolved build, player/saber alignment, and animation/audio issues while integrating systems across multiple gameplay maps.

Study Partner

Originated the idea and led project framing, resource selection, and task allocation for a five-person team. Co-developed a Meteor/React and MongoDB application to help students organize face-to-face study groups through course-based peer-support profiles, scheduling, and a shared calendar. The team documented feedback from five community members and built 12 automated end-to-end test scenarios, with GitHub Actions running ESLint and TestCafe.

Honors and certifications

Technical skills

Programming
Python, Java, C/C++, C#, TypeScript/JavaScript, SQL, Dart, HTML/CSS.
Machine learning and adaptation
PyTorch, PyTorch Lightning, Hugging Face Transformers, PEFT/LoRA, scikit-learn; bf16 training, gradient checkpointing, custom action-learning objectives.
LLM inference
vLLM, Ollama, bitsandbytes; 4-/8-bit quantization, tensor-parallel GPU inference, Gemini and OpenAI-compatible model APIs.
AI agents
Tool/function calling, JSON Schema validation, constrained decoding, multi-agent orchestration, tool/skill routing, memory and context management, hybrid local/cloud inference, failure-aware fallbacks.
Retrieval and grounding
Retrieval-augmented generation (RAG), multi-query evidence retrieval, query decomposition, structured context integration, source-grounded answer generation.
Vision and interpretability
CNNs, Vision Transformers, saliency and attention maps, Grad-CAM++, LIME.
Evaluation and verification
Agent benchmark design, paired counterfactual experiments, matched-budget and matched-seed ablations, deterministic replay, execution-trace analysis, exact policy enumeration, independent certificate checking.
Data and statistics
NumPy, pandas, SciPy, Matplotlib, seaborn; paired and cluster bootstrap, permutation/sign tests, sensitivity analysis.
Research infrastructure
Slurm, Hydra, Linux, Git; reproducible GPU experiments, checkpointing and resumable execution, data/model provenance, runtime validation.
Web and backend
Node.js, Angular, React, ASP.NET Core, Flask, Spring Boot, Meteor, Pydantic; REST APIs, WebSockets, SQLite, SQL Server, MongoDB; caching and request deduplication.
Mobile and real-time systems
Flutter, WebSockets, LAN client-server networking, Bluetooth Low Energy prototyping, Android permissions, camera/location APIs, SharedPreferences; Raspberry Pi and PLCs.
Game development and VR
Unreal Engine, Blueprints, UMG/Widget UI, Unity; VR controller interaction, gameplay state, materials, animation and audio integration.
Software testing and delivery
pytest, TestCafe, ESLint, GitHub Actions; automated end-to-end testing, continuous integration, DigitalOcean deployment.

Languages

English, Mandarin, Cantonese